From Nadir to Oblique

Path planning algorithms will bolster beyond visual line of sight UAV flights

While Canada and Europe are already relaxing their rules on flying UAVs outside the visual range of pilots, BVLOS (beyond visual line of sight) operations in the U.S. may still take some time for government approval due to airspace safety and security considerations. This is, of course, another classic case of technological innovation that’s out of step with bureaucratic rule making as more advanced algorithms for autonomous UAV navigation have now become more widely adopted and easier to deploy.

Take for example path planning, or the algorithms that permit UAVs to avoid obstacles and reach their target efficiently. Appropriate path planning is crucial so that UAVs operating under BVLOS can complete their flights safely while accommodating real time air traffic constraints.

Path planning makes this possible by finding the optimal path between the origin and destination points, while at the same time taking into consideration travel distance and time, as well as battery power. Crunching all this information can now be easily accomplished by the new generation of path planning algorithms.

In an in-depth analysis published in the Journal of Physics last year, Elena Politi and her colleagues from Harokopio University in Athens, Greece, identified three important factors that are shaping the development of safe and reliable BVLOS operations. They include path planning algorithms, sensors for environmental detection and perception (think of cameras, lidar, and radar), and satellite-based advanced surveillance technology that allows drones to determine their position in real-time (called ADS-B or Automatic Dependent Surveillance-Broadcast).

Path planning is so important to the UAV industry that many algorithms have already been developed throughout the years and drone users can choose algorithms to employ depending on the complexity of their BVLOS operations.

In general, path planning algorithms can be pigeon-holed as either global (users provide the path to be followed by the drone), local (real-time path planning using sensors), or a hybrid of both approaches. They have interesting names, such as the most commonly used A* algorithm, Ant, Dijkstra’s, Neural Network, Particle Swarm, and Genetic algorithms. But what unites all of them is their computational efficacy.

In a recent review of path-planning algorithms by researchers from Queensland University of Technology in Australia, global and local path-planning algorithms have been shown to exhibit various levels of efficacy. For example, algorithms such as the A*, Genetic, and Particle Swarm are more useful in “static and regulated situations,” meaning in environments where the obstacles are known beforehand, such as in cities and other built-up areas.

Conversely, local and hybrid path planning algorithms that are based on neural network or reinforcement learning have greater flying adaptability because they allow drones to make decisions in real time, although they require a lot of computing power.

Commercial UAV data platforms like DroneDeploy and Pix4D, are already offering these path planning options for their users. Yet according to an article in the journal Nature, available path planning features are often limited to only a few standard configurations. This means that flights, including those that operate under BVLOS, must capture their targets multiple times in order to perform their task effectively. This, according to the authors of the article, is a time-consuming process which can delay BVLOS projects.

Another setback to operating UAVs beyond visual line of sight, at least in the U.S., is acquiring a BVLOS waiver from the Federal Aviation Administration (FAA). The FAA oversees all drone operations in the country under its Part 107 rules, and procuring a Part 107 BVLOS waiver can be a complicated and arduous process. Since the rule was implemented in 2016, only a handful of companies have been granted waivers, mostly big drone delivery businesses such as Zipline and UPS Flight Forward.

So how about private drone operators who want to fly their machines for surveying and asset inspections? They could always apply to get the BVLOS waiver themselves, or they can use the services of companies like American Robotics and DroneDeploy, which have been granted BVLOS waivers by the FAA at both local and national levels.

These companies use advanced path planning algorithms that allow BVLOS-operated drones to access larger areas, reduce costs and security risks to operators, as well as expand to other tasks such as site inspections, mapping, and even sampling.

For now, all this talk of the benefits of BVLOS is limited to those who can acquire the FAA waivers. Whereas path planning algorithms are already proving their worth in normal UAV operations, drone operators who want to conduct BVLOS flights will just have to be patient for the release of a definitive set of BVLOS rules by the U.S. government.

In the meantime, other countries are on the path to launching uncrewed BVLOS drone operations. The UK, for example, will permit BVLOS flights by 2027, creating new jobs and bolstering the nation’s coffers up to £45 billion by 2030, according to estimates by PWC, a global consulting firm. In the high-stakes drone business, anything that promotes flight efficiency, safety, and cost savings is worth waiting.